1
\$\begingroup\$

I need to build a passive high-pass filter that has a cutoff frequency of 2-3 kHz. Sounds simple? It needs to pass between 10 and 20 W rms power at 20 kHz without overheating and going "bang".

The purpose of this circuit is to supply AC power to a nichrome wire heating element avoiding n*50 Hz amplifier noise due to mains electricity frequency. The rms voltage is below 20 V and the current is below 1 A.

In a previous attempt, I tried to build the circuit below, where R1 is the nichrome wire heater and C1 is two opposite 100 uF electrolytic capacitors in series (cathodes joined together):

enter image description here

This worked for about 10 seconds, then one of the capacitors exploded:

enter image description here

My questions are:

  • What do you think went wrong with my attempt? e.g. Was a 35 V capacitor rating not high enough, or was cooling needed?

  • What would you suggest will give the circuit the best chance of working as intended.

\$\endgroup\$
11
  • 2
    \$\begingroup\$ The caps must not be polarized nor high ESR in order to handle the high ripple current. THe wire type is also importance due to skin effect. WHat else is on the line? and why do you care about n*50Hz? \$\endgroup\$ Jun 4, 2018 at 20:50
  • \$\begingroup\$ Thanks for comment about ESR. If you or others can post a full answer about what that means in the context of this question, would be grateful. It is really beside the point of the question but if you want to know, the heater is used in a sensitive magnetic environment where stray fields with a frequency < 1 kHz must be minimized. \$\endgroup\$
    – MichaelT
    Jun 4, 2018 at 20:57
  • \$\begingroup\$ Unless you define the requirements better, any answer will be sub-optimal. Max magnetic field levels(f) and cable choices, source (impedance and frequency tolerance) Avoiding blowup is easy, Also budget or any other limitations. (cable length) Why not DC? (is that too easy?) \$\endgroup\$ Jun 4, 2018 at 21:05
  • \$\begingroup\$ I can give you what I know. Source is a 500 W audio amplifier rated to 25 kHz, output is 20 kHz sine at 15 V plus <50 mV n*50 Hz ripple that I want to get rid of. The load is nichrome wire 35 ohm, 32 gauge. The source and load are connected with BNC cables and the breadboard you see in the photo. Pretty much all I need is guidance on avoiding blowup. \$\endgroup\$
    – MichaelT
    Jun 4, 2018 at 21:10
  • \$\begingroup\$ DC also creates magnetic background that I want to avoid, so that method is ruled out. (few hundred nT, sounds small but is killer for me) \$\endgroup\$
    – MichaelT
    Jun 4, 2018 at 21:13

3 Answers 3

4
\$\begingroup\$

TL/DR: use a big film cap, like a motor cap.

DC also creates magnetic background that I want to avoid, so that method is ruled out.

Use a non-inductive pattern for your nichrome wire, this will cancel the magnetic field out.

What do you think went wrong with my attempt? e.g. Was a 35 V capacitor rating not high enough, or was cooling needed?

A 100µF general purpose cap has about 1-2 ohms ESR, which is resistance so it will convert current into heat. This creates enough heat to explode the capacitor. It is quite small, it can't dissipate too much heat.

You need a low-ESR cap. A good idea would be a cap specified for a ripple current higher than the AC current you run through it. (that's the idea with the ripple current spec).

Ceramic has voltage-dependent capacitance, but the voltage across the cap will be small. So, ceramic may be an option. You can use any parallel combination of caps optimized for lower cost, like 10x 10µF 25V caps. Be aware that it will add a bit of distortion on the waveform.

It will be tricky to find 100µF aluminium caps with low enough ESR and high enough ripple current, but you can use polymer caps, which should work well.

\$\endgroup\$
3
  • \$\begingroup\$ I am not worried about the size of the capacitors, polypropylene film looks like a good option to me. \$\endgroup\$
    – MichaelT
    Jun 5, 2018 at 6:19
  • \$\begingroup\$ @MichaelT yeah, plus the big motor caps are pretty cheap too (and very robust) \$\endgroup\$
    – bobflux
    Jun 5, 2018 at 17:31
  • \$\begingroup\$ A 1 uF polypropylene film capacitor worked for me. In the end a first-order filter (without the inductor) was sufficient. \$\endgroup\$
    – MichaelT
    Jun 8, 2018 at 18:17
3
\$\begingroup\$

This is the proper way to build a non-polarized capacitor out of ordinary electrolytics. The diodes are mandatory to protect each capacitor from reverse polarity. That is possibly what blew them before. Now the most reverse voltage across each one will be about -0.7 volts.

schematic

simulate this circuit – Schematic created using CircuitLab

\$\endgroup\$
2
  • \$\begingroup\$ Thanks, 1N4007 datasheet says maximum current rating is 1 A. I could put two in parallel with each capacitor to give a better safety margin. \$\endgroup\$
    – MichaelT
    Jun 5, 2018 at 6:21
  • \$\begingroup\$ Or use 1N5408 3A diodes. \$\endgroup\$
    – user105652
    Jun 5, 2018 at 6:25
1
\$\begingroup\$
Your filter:                My Filter:

Type: 2nd order HPF         3rd order HPF
Peak: 8 kHz +12 dB          +13 dB @20 kHz
Attenuation:   
   100 Hz:  -75 dB          -130 dB
  1000 Hz:  -35 dB           -70 dB

schematic

simulate this circuit – Schematic created using CircuitLab

ECWF4754JL  0.75 uF film PP radial ( 3pc ) 

The paired parts carry more than twice the current here at resonance thus shared current.

If you wind your own magnet wire choke, you can drive > 100 W rms or >200W peak in 4 Ohms.
30uH into 34 Ohms,
40uH into 4 Ohms for 20kHz enter image description here

Why under-design for 10~20W when you can over design for 100W?

\$\endgroup\$
1
  • \$\begingroup\$ I wondered about putting the resonance frequency at the frequency of my AC output (20 kHz). Are there any downsides of doing this? Temporal stability of the heater output is reduced? \$\endgroup\$
    – MichaelT
    Jun 5, 2018 at 6:29

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.